Every drug program has a breaking point
Conviction isn't evidence.We find what is.
Independent, AI-native scientific scrutiny for those who bet on drugs.
Request access →Independent, AI-native scientific scrutiny for those who bet on drugs.
Request access →You check Carfax before you buy a car.
When you bet on a drug, you check Prelucis.
It hides across decades of literature, thousands of trials, every comparable program and its fate, and the regulatory record — more than any team can connect by hand. Not a failure of skill. A failure of scale.
Once you're in, you've fallen in love — and the loudest conviction in the room becomes the house view. No one wants to be the one who says it breaks.
We read it all — at machine scale — then apply pressure until the fault lines show. We hand you what's underneath, before the catalyst does.
We don't ask whether it works.
We ask how it breaks.
Most of what a program reveals is scattered, buried, or never written where you would look. We read across all of it — continuously — and connect it into one picture.
Decades of literature — far more than any team can read by hand.
Every registered trial and its fate — including the ones that quietly stopped.
Approvals, rejections, and every concern raised along the way — worldwide.
What surfaces after launch, where programs break in the open.
Where a mechanism is first stress-tested — and often first fails.
The failure record, and signals that sit outside the obvious sources.
Not more data — the sources no one connects, read together and put under pressure.
The record is built on success — wins get published, failure stays buried and confidential. But failure is where the signal lives, and it vastly outnumbers the wins. We start there: surfacing what failure leaves behind and folding in the rest, to rebuild a picture more complete than the one you were told.
We ingest everything above — and more — far past what any desk could read by hand, then connect it into one picture.
We don’t look for reasons it works. We attack it — where the endpoint bends, where the mechanism won’t travel, where comparable programs died.
Every fault line carries its provenance and a calibrated confidence — each stated on its own, never rolled into one number.
Responder-subgroup sensitivity collapses once placebo drift is normalised. The registered analysis masks it.
Unsupported beyond the lead indication. Three analogous programs failed on the same axis.
It assumes a surrogate endpoint the agency has twice declined to accept.
Prelucis is in private release with a small number of funds.